Development of High Quantum Efficiency UV/Blue Photocathode Epitaxial Semiconductor Heterostructures for Scintillation and Cherenkov Radiation Detection

13 Pages·2002·0.7 MB·English

by  Leopold, Daniel J

Development of High Quantum Efficiency UV/Blue Photocathode Epitaxial Semiconductor
Heterostructures for Scintillation and Cherenkov Radiation Detection
Summary of Research
Principal Investigator: Daniel J. Leopold
7-15-00 to 7-14-02
Washington University
Department of Physics
St. Louis, Mo 63130
NASA Grant Number: NAG5-8536
Overview,Significance,andImpact of Research
Enormous technological breakthroughs have been made in a variety of optoelectronic
devices through the use of advanced heteroepitaxial-semiconductor crystal growth techniques
such as molecular beam epitaxy (MBE). The primary goal of this research project was to further
extend this technology and to demonstrate significant gains in UVfolue photonic detection by
designing and fabricating atomically-tailored heteroepitaxial GaA1N/GaInN photocathode device
structures. Although a simple activated p-type single layer GaN structure can deliver 10 to 20%
quantum efficiency, a more complex design is required to achieve the higher levels that our
research is targeting. Band gap engineering concepts are being employed in the photocathode
structures that we have fabricated by MBE. The high level of sophistication and complexity of
our approach required some further effort in the basic optimization before completed devices
could be evaluated. No other approach has the same potential to reach the high quantum
efficiencies (in excess of 50%) that we are targeting in the UV/blue waveband.
This NASA Explorer technology research program has focused on the development of
photocathodes for Cherenkov and scintillation radiation detection. Support from the program
allowed us to enhance our MBE system to include a nitrogen plasma source and a magnetic
bearing turbomolecular pump for delivery and removal of high purity atomic nitrogen during
GaA1N/GalnN film growth. Under this program we have also designed, built and incorporated a
cesium activation stage. In addition, a connected UHV chamber with photocathode
transfer/positioner components as well as a hybrid phototube stage was designed and built to
make in-situ quantum efficiency measurements without ever having to remove the photocathodes
from UHV conditions. Thus we have constructed a system with the capability to couple
atomically-tailored MBE-grown photocathode heterostructures with real high gain readout
devices for single photon detection evaluation. Complete GaN/InGaN photocathodes have been
designed and fabricated on up to 2-inch diameter sapphire substrates by MBE (Fig. 1). The
structural, optical, and electronic properties of these films have indicated very good quality
material. We have activated photocathode surfaces with cesium and measured the photoelectron
emission quantum efficiency and spectral response.
Figure 1: A two inch diameter graded GaN/InGaN wafer grown by molecular beam epitaxy.
Byalloyingwith In, wehaveextendedtheresponseofGaNphotocathodesintothevisible
band.Enhancedsensitivitytobluelightshouldenableanewgenerationofphotomultipliertubes
for scintillation light detectionin X-ray detectors,activeshieldsfor X-ray experiments,and
scintillating fiber detectors.Optical monitorsfabricatedusing suchphotocathodesarelikely
componentsin future wide field X-ray experiments. High efficiency photocathodeswith
extended wavelength coverage, based on AIGaN/InGaN are expected to supplement
developmentsin high-resolutionreadoutsand will provide a powerful elementof future
generationsofUV instruments.
Photocathode Design, Epitaxial Growth, and Material Quality
Our research on high quantum efficiency photocathodes involved the design and fabrication
of precisely-tailored heteroepitaxial semiconductor structures that have sensitivity in the UV/blue
spectral range. We selected the AlGaN/InGaN system mainly because the band gap can be
tailored over an energy range from 1.9 to 6.2 eV and epitaxial thin film layers can be grown
directly on optically transparent sapphire substrates. All of the A1N/GaN/InGaN heterostructures
discussed in our work have been fabricated by MBE.
The use of MBE for crystal growth makes it possible to control film composition on an
atomic scale and to fabricate abrupt heteroepitaxial interfaces. This ultra-high-vacuum technique
allows for very quick on/off switching of atomic and molecular beams through the use of shutters
in front of each thermal or electron beam source. The system we are using for GaA1N/GalnN is
equipped with an RF plasma source to deliver atomic nitrogen to the substrate for growth, as well
as an oil-free magnetic bearing turbomolecular pump for removing excess molecular nitrogen
from the ultra-high vacuum chamber. Standard MBE effusion sources are used for both Ga and
In as well as the Mg p-type dopant. Electron beam sources are used for the A1 flux and Si n-type
dopant.
A reflection high-energy electron diffraction (RHEED) system mounted inside the vacuum
chamber allows the surface crystal quality to be monitored during growth as individual atomic
layers are added to the surface one at a time by examining surface reconstruction diffraction
patterns. This feedback provides the absolute finest control of the heteroepitaxial crystal growth
process, thereby making possible precise fabrication of semiconductor layered structures for
fabricating high quantum efficiency photocathode devices. Figure 2 shows typical RHEED
patterns from GaN grown on sapphire substrates in our MBE system. The streaked pattern
shown on the left indicates an atomically-smooth growth front. We have found that the RHEED
pattern is very sensitive to the ratio of Ga to N flux during growth in that Ga-rich conditions lead
to a highly streaked diffraction pattern while N-rich conditions produce a more spotted diffraction
pattern as shown on the right side of Fig. 2. However a high flux ratio can also lead to Ga
precipitation on the surface. The best films are produced when the ratio of Ga to N flux is close
to one.
Figure 2: Left: RHEED pattern of GaN captured during the growth process with Ga/N flux ratio near
one. Right: RHEED pattern for a GaN layer later during growth with Ga/N flux less than one.
The potential to increase the quantum efficiency of these photocathodes relies on a couple of
key design features that can be incorporated in our heteroepitaxial layers. Examples of this are
shown in Fig. 3 where conduction and valence band edge energy spatial profiles are displayed for
two possible photocathode designs. This diagram is a rough schematic plot of band gap versus
depth into the photocathode structure. For clarity the thickness of individual layers shown in
Fig.3 are not drawn to scale.
AIN
$ubstr_e Vacuum
F.,
[nGaN:Id|
Energy
C.zO
*,-dv,n..A
E,
I.
AIN
Irld_
Vacuum
-- IRGtN: MI
Em'ru
i )
Figure 3: Spatial profiles of conduction and valence band edge energies, Ecand Ev respectively, for
composition variations versus depth in two different photocathode heterostructures shown at the top
and bottom. Evac is the vacuum energy near the photocathode surface.
All of the GaN/GalnN heterostructures used in our studies were grown on single-crystal
sapphire substrates. The mechanical strength and UV/visible optical transparency properties of
sapphire make it an excellent choice as a window material for photocathode structures. An AIN
optical antireflection layer can be inserted between the sapphire and the A1GaN/GalnN
photocathode region, as shown in Fig. 3, to serve as a wide-band-gap barrier to prevent electronic
back diffusion into the substrate interfacial region where defect densities are expected to be
higher and nonradiative recombination of photoexcited electrons larger. Inserting this wide band
gap AIN buffer layer in the structure ensures that photoexcited electrons do not diffuse back
toward the sapphire substrate interface, but instead are reflected toward the photocathode
emission surface. We can also grade the alloy composition in the photocathode region to tailor
the conduction and valence band profiles in a manner which will enhance photoelectron transport
to the surface. Examples of this are shown in Fig. 3, where conduction and valence band-edge
energy spatial profiles, Ec and Ev respectively, are displayed for two photocathode designs. In
Fig. 3 the GalnN alloy composition spatial profile is graded not only to optimize the electronic
transport to the front surface but also to extend the optical absorption to longer wavelength. The
energy gap of GaN in the wurtzite crystal structure is 3.39 eV at room temperature [1]. This can
be adjusted to lower energies through the addition of In to the alloy
It is known that an electric field applied inside a semiconductor photocathode layer drives
electrons toward the emitting surface and in so doing can increase the quantum efficiency by as
much as a factor of two [2]. As shown at the top of Fig. 3 we have been exploring a design to
create these internal fields in the photocathode layer by grading the alloy composition, which tilts
the conduction and valence band edges. As the In content in the alloy layer increases the energy
gap between valence band and conduction band decreases, resulting in a sloping of the band
edges. Although the fractional amount of change in the conduction and valence bands are
different, the overall effect is to tilt the conduction band since the p-type dopant incorporated
throughout the layer allows mobile hole charge carriers to diffuse in a manner that minimizes the
energy, leaving the valence band profile flat. The tilted conduction band drives photoelectrons
toward the surface, increasing their escape probability and thus the quantum efficiency. Finally,
an activation layer of Cs on the surface bends the bands to achieve a negative electron affinity
(NEA) condition, which is vital for having a high photoelectron escape probability. We have also
been considering novel means of activating the photocathode emitting surface in order to achieve
the NEA condition. Since AIN and A1GaN with high A1 concentration have an intrinsic NEA
surface, we suggest the possibility of ending the epitaxial layers with a Si-doped A1GaN surface
in order to achieve an NEA condition without the need for post-growth Cs activation. This design
is shown at the bottom of Fig. 3.
Adding In as an alloy component to the GaN epitaxial layer does induce strain due to the
lattice mismatch between GaN and InN. It can also lead to elemental phase segregation in the
alloy layer. This puts some limits on the epitaxy and also narrows the growth parameter window.
In spite of these constraints we have come up with a method that appears to work and allows
grading of the alloy with good crystal quality. X-ray diffraction and TEM lattice image studies
show good registry of epitaxial GaN and InGaN layers with the c-plane oriented, single-crystal
sapphire substrates (see Figs.4, 5, and 6).
_.5x10 5 _ ,
_.OxlO 5
GaN
3" Sapphire
Sapphire
;.5xtO 5
.OxtO 5
v
_.OxlO 4
GaN
L__
0
-_,OxtO 4
I I ,ll
20 40 60 80 100
Figure 4: X-ray diffraction 0-2_ scan showing peaks froa_({:_eg_GaNepitaxial layer grown on a
single-crystal c-plane sapphire _ubstrate.
0•'0b
'd I
z J
t¸ _.[]- 10
,}:j 3,,t 'j_ 3,b
2:: (_i%(_.)
Figure 5: X-ray diffraction O-2Oscan and Voigt function fit for an epitaxial GaN/InGaN heterostructure
showing a double peak structure, indicative of the c-plane lattice spacing differences between the two
layers resulting from lattice expansion in the alloy layer.
Figure6: TransmissioenlectronmicroscopcyrosssectionailmageofaGaN/InGaNstructuregrownona
sapphiresubstratsehowingparallecl -planeorientedlatticeplanes.
Wehavealsodevelopeda methodto fit x-ray peaksusingVoigt functions. By removing
instrumentalbroadeningcontributionswe canevaluatedefectdensitycorrelationswith x-ray
peak widths. We can also fit several peaks arising from multilayered GaN/InGaN
heterostructurestoextractlatticestrainprofiles(Fig.5). TheFWHM of theGaN(0002)X-ray
peakis approximately400arcsecondsfor a0.5pm-thickfilm, whichindicateshigh-qualityfor
suchthinepitaxialGaNonsapphire. Optical absorption measurements of our GaN and InGaN
samples confirm the expected band gap shift with increasing In concentration as shown in Fig. 7.
I00_
8O
v
6O
e_
"GaInN (low In) ,"
4O
E--
!
20 ,J ,," GalnN (high In)
i
0 ,,, .... _%1.... i, i........
2OO 300 400 500 600 700
Wavelength (rim)
Figure 7: Optical transmission as a function of wavelength for GaN and InGaN samples showing the
band-edge shifting at higher In concentrations.
7
Another very important component of our photocathode development has been the ability to
achieve high p-type doping levels in the active InGaN layer. Typically Mg is used as a p-type
dopant in GaN. It is difficult to get highly p-type GaN epitaxial material in part due to the large
intrinsic background n-type carrier concentration, as well as the large ionization energy of the Mg
p-type acceptor level [3]. We had success with Mg by first reducing the n-type background by
growing under slightly N-rich conditions and also by growing on the Ga-face, which is achieved
by exposing the sapphire surface to a nitrogen plasma prior to growth [4]. Hall measurements,
using the Van der Pauw technique, resulted in p-type carrier concentrations near 10_8cm -3. We
found this to be adequate for fabricating semiconducting photocathode layers with negative
electron affinity (NEA) surfaces.
Photocathode Quantum Efficiency
During this research program we have developed a sophisticated system for determining the
absolute photon conversion efficiency and spectral response of our photocathode detectors using
a photon counting system and differential measurements with respect to a calibrated absolute
reference. This system is not sensitive to intrinsic gain of the photocathode or to other systematic
uncertainties such as lamp drift, pedestal variations, alignment or charge-collection efficiency.
As mentioned earlier, the MBE system is comprised of three primary components: the
growth chamber, sample introduction and Cesium activation chamber, and the quantum
efficiency chamber (see Fig. 8). The photocathode wafer can be transferred from the growth
chamber to a carousel for activation in the sample introduction chamber. Thereafter, the carousel
can be translated back and forth between the activation chamber and the quantum efficiency
chamber. This system makes possible in-situ measurements of quantum efficiency and spectral
response without breaking vacuum, providing the best feedback for optimizing wafer growth and
surface activation.
Figure 8. Front and back views of our custom-built photocathode quantum efficiency measurement
system connected to the MBE vacuum chamber.
A monochromaticp, ulsedlight sourcehasbeeninstalledonasmallopticalbenchnearthe
quantumefficiencychamber(seeFig.8). Theopticalsystemconsistsof apulsedXeflash-lamp
followedby amonochromatorundercomputercontrol. A custom-madebifurcatedfiber bundle
consistingof 14interleavedUV transparenftused-silicafibersisformattedontotheoutputslit of
themonochromatorto evenlysplit thelight into two pathseachconsistingof a close-packed
bundleof 7 fibers. One fiber bundledelivers light to a UV-enhancedsilicon photodiode
referencedetectorlocatedin ashieldedenclosureontheopticalbench. Theotherbranchofthe
fiber bundlegoesto a focusingbeamprobeheldby anXY positionerabovea windowof the
vacuumchamber. Light is focusedononeendof a fused-silicafiber mountedin thevacuum
chamber.Thisfiberisterminatedwithasmallcollimatinglensthatilluminatesthephotocathode
surface(Fig.9).
Figure9. Viewsinsidethevacuumchambeorfthequantumefficiencymeasuremesnttation(left)
showingthehybridphototubehousingandphotocathodweafercarousela,ndtheCesiumsourcefor
activatingphotocathodseurface(sright).
TheXYZ/anglepositionerbuiltinto theQEchamberallowsthephotocathodeto bemated
to anelectronmultiplierdevicein thevacuumchamber(Fig.9). Thephotocathodeispositioned
at the entrancewindow ota barehybrid photomultiplieranodestructure. A high voltage
feedthroughprovidesa0-10kV biasbetweenthephotocathodecarouselanda segmentedarray
of sevenphotodiodes. (This device is a windowlessversion of the DEP P25 HPMT
manufacturedby Delft Instruments.) Photoelectronsemittedby the cathodeareaccelerated
towardthephotodiodestoanenergyof 10keV,liberating-2000 electrons in the photodiode. In
this proximity-focused device, the field is sufficiently large that the electrons travel in a straight
line from the photocathode to the corresponding segment of the photodiode array. The
arrangement of photodiodes is such that a single central pixel is surrounded by a guard ring of six
hexagonal-close-packed detectors. Signals from these photodiodes are read out through another
vacuum feedthrough to the external preamplifiers. Hybrid photomultiplier tubes (HPMT) of this
variety are ideally suited for efficiency measurements. The gain is linearly dependent on the bias
voltage, unlike the dynode chains of photomultiplier tubes. These devices also have the potential
9
to provideexcellentsinglephotoelectronresolutionthatallowsanabsolutegaincalibrationby
identifying,e.g.,theone,twoandthree-photoelectropneaks.
The signalsfromthereferencesilicon detectorandtheHPMT detectorarereadout with
AMPTEX 250 low-noisepreamplifierswith input JFETsmatchedto the capacitanceof the
silicon detectors. Thesetwo signalsarethenfed to shapingamplifiersand analog to digital
converters. A LabView application controls the light pulser, scans the monochromator, and
simultaneously captures the waveforms from signals in the calibrated reference detector and from
the photocathode/HPMT. Capturing the triggered waveforms allows for real-time pedestal
subtraction, and provides an evaluation of electronic noise not available in electrometer or lock-in
systems. Like a lock-in, the light pulses can be averaged over the appropriate time-window
(following the trigger) increasing the signal to noise ratio with acquisition time. The relative
measurement corrects for drift or fluctuations in the light pulser. All of these features combine to
provide an excellent handle on systematics in the relative measurement. Figure 10 shows single-
shot pulses from a GaN/InGaN:Mg photocathode and from the reference detector obtained with
our system for very low lighl levels using a neutral density filter (ND = 2) in the fiber optic beam
line.
CIAmpl l
/ i : : 2.52 v
.... _i.... _:.... :?.... i:......].. _:......].... :2,40V "
i....t.!.i..;...i....:. :...i....!....i.,.
.... _ii C_ Amp
: Refc-n_nce(CO •
•., .......... :....... :ft ......... :.... :....
'li. i : i l\ i ! i
ChLi,,?'.09...V... 9_'.. J,'_V..... M2o_........
Figure 10: ln-situ single-shot measurement of photoelectron emission from an MBE-grown GaN/InGaN
photocathode compared to a reference detector in response to pulsed UV light excitation at 300 nm.
The system is calibrated for absolute measurements by placing a calibrated photodiode
(identical to the reference device) inside the vacuum chamber and performing a cross calibration.
This procedure was used to renormalize the measurements in the reference arm to provide an
absolute measurement of the number of photons per light pulse hitting the photocathode.
Together with the HPMT gain calibration, this provided a complete absolute calibration of the
system.
Traditionally quantum-efficiency has been measured by illuminating a cathode in a simple
tube-diode structure to determine the photocathode current for an applied light level. Quantum
efficiency is defined in terms of such measurements, as the ratio of the number of electrons
collected on the anode to the number of incident photons. However, intrinsic gain in devices can
give rise to an apparent quantum efficiency in excess of the actual photon detection efficiency
Our system has the potential to distinguish between the single photon detection efficiency and the
quantum efficiency.
10